Improvement Efficiency of Radial Basis Function Based on the Optimization of its Parameters using Particle Swarm Optimization

Malekpour Golsefidi, Majid; Ali Abbaspour, Rahim

doi:10.22059/eoge.2023.351405.1126

Improvement Efficiency of Radial Basis Function Based on the Optimization of its Parameters using Particle Swarm Optimization

Document Type : Original Article

Authors

Majid Malekpour Golsefidi

Rahim Ali Abbaspour

University of Tehran

10.22059/eoge.2023.351405.1126

Abstract

One of the most widely used interpolation methods is the application of radial basis
functions (RBF) to achieve a global and exact surface. Because of the computational
complexity and fluctuation of the fitted surface achieved by the RBF method, the
interpolation method is converted to a radial basis function neural network (RBFNN)
model to solve these problems. Particle swarm optimization (PSO) algorithm is used in
the designing process of a neural network to determine the optimal values of the center and radius of each basis function. In addition, the weights of radial basis functions are
determined by calculating the pseudo-inverse matrix of coefficients. Finally, the
accuracy of the proposed method was evaluated using the root mean square error (RMSE)
in areas with different elevation ranges. Consequently, the proper type of the radial
basis function was selected based on the RMSE in each area. As a result of the study,
RBFNN model has a higher accuracy compared to other interpolation methods
especially the RBF interpolation method.

Keywords

Radial basis function

neural network

interpolation

particle swarm optimization

clustering

Subjects